Minimal weights covered tree on Level Set segmentation using local curvature constraints

Owing to the inhomogeneity and ill defined edges present in images, the incorporation of prior knowledge into level set models, in image segmentation, is a field of active researches. In this paper, a new way of incorporating prior information to constrain the evolution of the level set model during the segmentation is presented. This technique allows resolving the problem of applying the same curvature coefficient in all image regions. We construct, based on Kruskal algorithm, the minimal weights covered tree of the initial density graph due to boundary curvature. The simulation results using different kind of images show that we get better results with respect to propagation, precision and homogeneity between the final propagating contour and local regions, compared to the classical level set method.